Development of an Address Geocoding System in Tokyo in the Meiji Era

Abstract

In recent years, historical geographers have used geographic information systems (GIS) to develop historical databases. However, researchers still lack systems that can handle large amounts of historical geographic data because preparing these databases is time-consuming. Address geocoding systems that convert addresses into coordinates (latitude/longitude) have been developed; however, most of these systems in Japan can only convert present-day addresses and cannot convert historical addresses of the Meiji and Taisho Eras. This study aimed to develop an address geocoding system that can effectively convert historical addresses from old Tokyo’s 15 wards in the Meiji Era into coordinates. To carry out this study, we used printed maps of 1907 to construct databases of historical addresses and developed an algorithm to search the databases. We examined 15 wards that existed for 44 years (1889–1932) in old Tokyo. Our databases contained attribute information such as the historical address strings and their coordinates. We obtained this data using GIS to digitize polygon data from geometrically collected map images. Because the coordinates of each address had to be inside the polygon, we made points within the polygon and used them as the representative points of each address. After we entered an address into the system, the algorithm searched the databases by matching the characters in each address string, starting with the beginning of the string, until it found a mismatched character. The system replaced each mismatch with the special character “_” and searched until no result was returned. Because every address string that we entered matched a string in the databases, we were able to find the correct addresses and their coordinates. Although converting addresses into coordinates using our system takes more time than existing geocoding systems for present-day addresses, this did not pose a problem. Our comparisons of conversion accuracy over different publication dates of addresses indicate that our system accurately handles historical address data from the 1890s to the 1920s.